Professional Documents
Culture Documents
Structures: Ahmad Siar Mahmood Shah, Saber Moradi T
Structures: Ahmad Siar Mahmood Shah, Saber Moradi T
Structures: Ahmad Siar Mahmood Shah, Saber Moradi T
Structures
journal homepage: www.elsevier.com/locate/structures
A R T I C LE I N FO A B S T R A C T
Keywords: Steel plate devices are commonly used for the seismic energy dissipation and damage mitigation of structures.
Steel plate fuse Particularly, butterfly-shaped steel plate fuses are essential components in controlled rocking frames, which have
Energy dissipating device been shown as promising self-centering systems for minimizing financial losses due to repair costs and downtime
Sensitivity analysis in strong earthquakes. The cyclic behavior of a steel plate fuse varies depending on design factors related to its
Statistical design of experiment
geometry and material. This paper aims at assessing the effect of seven factors on the cyclic response of steel
Finite element simulation
plate fuses. These factors include the fuse length, mid-width, end-width, thickness, band zone length, fuse yield
strength, and modulus of elasticity. Detailed finite element models are first developed and validated against past
experimental test results. A design of experiment method is used to statistically evaluate the effects of the design
factors and their interactions on the cyclic response characteristics of steel plate fuses in terms of initial stiffness,
yield strength, ultimate stiffness, effective damping, maximum strength, and ductility. The results show that the
most significant effects are the fuse end-width, thickness, and length.
1. Introduction previously used solid steel plate fuses [4]. The openings divide the plate
into several parallel links, thereby making the energy dissipating plate
The conventional approach to seismic design is based on ductile more flexible. As a result, its behavior is mainly dominated by flexural
behavior of structures. Appropriate strength and inherent ductility of mode behavior compared to predominant shear mode behavior of solid
structures allow inelastic drift capacity and energy dissipation during steel plates without openings [4].
severe seismic motions. Steel plates with engineered cuts can be im- Previous research has been conducted to improve shear deformation
plemented in the structures as the sacrificial ductile structural elements capacity through improving the buckling resistance of the shear plates
and easily replaceable fuses. The replaceability of these plates are re- by means of controlling compactness (the width-to-thickness ratio) of
ferred to as structural fuses, which experience damage and thus limit each single link compared to the original solid plate [4] (Fig. 1(b)).
the seismic forces applied to the structure [1,2]. By confining the in- Relatively, steel slit fuses, which mostly possess a link width-to-thick-
elastic deformation and structural damage to steel plates with shear ness ratio between 1 and 2, exhibit stable hysteretic behavior up to a
links, the primary structural elements like beams and columns remain peak shear deformation between 10% and 20% [5,6]. Along the same
essentially elastic and undamaged. One example of using structural lines, supplying additional shear deformation capacity for structural
steel plate fuses is in controlled rocking steel braced frames, which have systems is relied on the out-of-plane buckling of the damper [7]. In a
been shown as promising self-centering systems that can minimize fi- study by Chan et al. [6], steel plate fuses fractured after reaching their
nancial losses due to repair costs and downtime in structures under peak shear deformation at an ultimate strength level twice the amount
strong earthquakes [3]. of their yield strength [6].
Extensive research has been conducted on different configurations The alteration of the straight openings with varying mid-width
of steel plates to investigate their ductility and energy dissipation honeycomb-shaped openings, leaving butterfly-shaped links, affects the
capability. Different shapes and dimensions of steel plate fuses result in yielding capacity, shear deformation, strength, and stiffness of the fuse
distinct cyclic response characteristics, such as shear deformation and [8,9]. The resemblance of butterfly-shaped links (Fig. 2) to the moment
ductility. The main goal in all varied configurations is to achieve a diagram of beams deforming in double curvature results in a wider
higher ductility considering the simplicity and cost-effectiveness of yielding distribution along the length of the plate [4,9]. Along the same
steel plates [4]. Fig. 1(a) illustrates an example of a steel plate with lines, in the experimental tests by Kobori et al. [9], butterfly-shaped
straight cut openings or slits, which shows higher ductility among all fuses exhibited 30% shear deformation without any stiffness
⁎
Corresponding author.
E-mail addresses: ahmad.mahmood@ryerson.ca (A. Siar Mahmood Shah), s.moradi@ryerson.ca (S. Moradi).
https://doi.org/10.1016/j.istruc.2019.12.026
Received 18 September 2019; Received in revised form 3 December 2019; Accepted 23 December 2019
2352-0124/ © 2019 Institution of Structural Engineers. Published by Elsevier Ltd. All rights reserved.
A. Siar Mahmood Shah and S. Moradi Structures 23 (2020) 799–811
Fig. 1. (a) Steel plate with slits or straight cut openings and (b) details of a single link.
800
A. Siar Mahmood Shah and S. Moradi Structures 23 (2020) 799–811
2. Motivation and objective of this study butterfly-shaped fuses), as shown in Figs. 3 and 4, respectively. Table 1
summarizes the details and dimensions of steel plate fuses used for
In this paper, the sensitivity of the cyclic response of steel plate fuses verification purposes. In Table 1, r denotes the radius of the cut at the
to different design parameters is assessed. In previous studies (e.g., end of the link and n is the number of the links in the specimen.
[4,27]) the variability of the fuse cyclic response to a number of in- For a satisfactory level of accuracy in the finite element simulations
dividual parameters and their ratios such as thickness, end-width to of highly nonlinear steel plate fuses, several trial models were analyzed.
thickness, length to thickness, mid-width to end-width of steel plate To reduce computational cost (i.e. running time duration and demand
fuses has been addressed. However, there is no previous statistical study for resources), only one link of the fuse specimen is modeled excluding
of main factor and interaction effects on the fuse cyclic response. In the experimental setup components. Proper boundary conditions are di-
present paper, a statistical design of experiment method is used to ex- rectly applied to the fuse model to simulate the accurate behavior of the
amine the effect of each parameter by changing different design para- fuse. Initial trial runs were also explored with different material prop-
meters together. Furthermore, possible interactions between design erties and element types.
parameters are determined. The presence of an interaction between two In the test setup by Lee et al. [22], the bottom of the strip damper is
factors is statistically indicated when the response sensitivity for one sandwiched between two thick angles that are bolted to a rigid reaction
factor varies at different levels of another factor. beam. This is simulated in the model by fixing all the degrees of
In this study, first, a detailed finite element model is generated to freedom at the bottom of the plate at the bolt line. At the very top of the
accurately capture the response of steel plate specimens against past test setup, the loading beam is constrained against out-of-plane trans-
experimental and numerical studies [22,28]. Following that, a sensi- lation at all locations and it is supported by two vertical struts on two
tivity study is performed using a two-level factorial design with seven sides of the loading beam. The loading beam is bolted to the top edge of
factors that potentially influence the fuse cyclic response character- the plate where it is constrained against out-of-plane as well as vertical
istics. The selected factors include design factors related to both the fuse translation. This is simulated in the finite element model by assuming a
geometry and material. The geometry related factors include the fuse roller reaction at the top edge of the specimen where it can only move
link length, mid-width, end-width, thickness, and band zone length. in the direction of loading and is constrained against out-of-plane and
The effect of the fuse yield strength and modulus of elasticity as ma- vertical translation. In the test setup, the load is applied at the top edge
terial-related factors are also evaluated. Thirty-two finite element of the loading beam whereas in the finite element models, the load is
models of fuses with different design factor combinations are analyzed; applied at the top edge of the plate. It is also worth mentioning that the
and the cyclic force-displacement response of each steel plate fuse is band zones at the top and bottom of the plate are modeled to better
assessed in terms of initial stiffness, yield strength, ultimate stiffness, capture the stiffness of the model. The finite element model is shown in
effective damping, maximum strength and ductility. Fig. 5.
SHELL181 elements with reduced integration are used for the
models in this study. This element type is suitable for analyzing thin to
3. Finite element model moderately thick shell structures. SHELL181 is a four-node element
with six degrees of freedom at each node: 3 translational (Ux, Uy, Uz)
To assess the cyclic response sensitivity of steel plate fuses, 3D finite and 3 rotational (ROTX, ROTY, ROTZ) degrees of freedom. From the
element models are developed and analyzed in the finite element ANSYS material library, several types of material models are compared
software, ANSYS Mechanical APDL [29]. Two specimens from the to check the behavior of the model. The hysteretic curves are then
uniform strip steel dampers tested by Lee et al. [22] are chosen for the compared with the experimental results to find the best matching ma-
calibration and verification of finite element modeling results under full terial model for the numerical simulations. Different material models
cyclic loading. Further, three different single-link specimens from the were used, including bilinear elastic perfectly plastic, multilinear ki-
numerical study by Farzampour and Eatherton [28,32] are modeled to nematic hardening model, the NLISO model (Chaboche model com-
verify the accuracy of the finite element analysis results in this study, bined with nonlinear isotropic hardening) and combined isotropic and
which is based on the response of a single link of the fuse. These spe- kinematic hardening model. As a result of these comparisons, it was
cimens exhibit three different limit states, including flexural yielding, concluded to select NLISO model (Chaboche model combined with non-
shear yielding, and lateral torsional buckling. linear isotropic hardening) as the material model. An elastic modulus of
The experimentally tested specimens by Lee et al. [22] include a 200 GPa and yield stress of 295 MPa are used for steel based on the
conventional prismatic strip damper (PSD, which has slit or straight coupon tests reported in Lee et al. [22].
openings) and a dumbbell-shaped damper (DSD also known as
Fig. 3. Details of specimen PSD-5 used for the finite element analysis verification (all dimensions are in mm) [22].
801
A. Siar Mahmood Shah and S. Moradi Structures 23 (2020) 799–811
Fig. 4. Details of specimen DSD-5 used for verification purposes (all dimensions are in mm) [22].
Table 1 important to create fine enough and regular meshing in the model for
Detailed dimensions of the specimens used for verification purposes [22,28]. capturing accurate cyclic response and overcoming convergence pro-
Specimen L (mm) a (mm) b (mm) t (mm) r (mm) n blems.
An element size of 3.5 mm is used for the fuse model based on a
PSD-5 180 36 36 10 5 10 mesh sensitivity study. This mesh size will result in having 6, 20 and 46
DSD-5 180 18 36 10 5 10 elements over the mid-width, end-width and length, respectively, fol-
Specimen with shear yielding 1000 100 500 20 5 1
lowing the recommendations by Farzampour and Eatherton [28,32].
Specimen with flexural 1000 100 300 20 5 1
yielding The finite element model for specimen DSD-5 for instance, consists of
Specimen with lateral torsional 500 100 100 16.67 5 1 108 key-points, 152 lines, 40 areas, 6662 nodes, 5944 elements (Fig. 6).
buckling It is confirmed that imperfection is not significantly influencing the
response of steel fuse models. Therefore, no imperfection is introduced
in the models.
The NLISO material model needs the specification of three different Displacement-controlled loading is used in the analyses. Similar to
constants, including; R0 , R∞, and b0 . These values characterize the the reference experimental study [22], the loading protocol consists of
isotropic hardening behavior of the material. R0 and R∞ are the initials symmetrical cyclic drift ratio with incrementally increasing amplitudes.
isotropic hardening modulus, R0 is the asymptotic hardening, and b0 The amplitude of each single step and the number of cycles for each
controls the decrease of hardening in the function of equivalent plastic step allow the specimen to undergo a large range of cumulative plastic
strain. For the verified models, values of 16, 30 and 50 are used for R0 , deformation [22]. The amplitudes of the loading for the first 12 steps is
R∞, and b0 , respectively [30]. increased with a ratio of 1.40. For instance, the 1st load step is set to
After defining the element type and material properties, mesh 0.44 mm and the 2nd load step is obtained by the multiplication of 1.40
control features in ANSYS Mechanical APDL [29] are used to control and 0.44 (i.e., the second load step = 1.40 × 0.44 mm = 0.62 mm).
the mesh density. These features include assigning line and element Following the 12th step, the increments have been fixed at 5.40 mm.
sizes. An automated free mesh, which gives the most desirable regular Fig. 7 shows the loading program.
meshing in irregular shapes is generated for the entire model. Fig. 6 Nonlinear static analyses with full Newton-Raphson solution
shows the meshing for the finite element model of specimen DSD-5. It is scheme are performed in ANSYS [29]. Geometric and material
802
A. Siar Mahmood Shah and S. Moradi Structures 23 (2020) 799–811
nonlinearities are included in the displacement-control analysis. In Note that the reference study is also based on validations against ex-
order to avoid convergence difficulties, the minimum number of sub- periments [26]. The lateral load-displacement plots of these three
steps to be taken is set to 10 and 25 for small and large amplitude load specimens are compared and shown in Fig. 9.
steps, respectively.
3.2. Sensitivity analysis
3.1. Validation of finite element analysis
A sensitivity analysis of the cyclic load-displacement response of
This section presents the validation of the finite element modeling steel plate fuses is conducted to assess the effects of different varying
for different specimens, including two steel plate fuses (DSD-5 and PSD- design parameters based on the statistical design of experiment method
5) among those specimens experimentally tested by Lee et al. [22] and and the validated finite element models. Seven potentially influential
three single-link specimen models each with different governing limit factors on the cyclic response characteristics are considered. Five of the
states [28,32]. Fig. 8 shows the lateral load versus lateral displacement factors are related to the fuse geometry, including the fuse length, L ;
of specimens DSD-5 and PSD-5 in comparison with the experimental mid-width, a ; end-width, b ; thickness, t ; and band zone length, c . The
response curves. As shown in Fig. 8, the numerical results are in good remaining two factors are material-related factors, including steel yield
agreement with the cyclic test results. The error in predicting the peak strength, Fy and modulus of elasticity, E . Two levels are considered for
strength using the finite element model is 1.14% and 3.50% for speci- the input factors. Table 2 lists these factors along with their low (−)
mens PSD-5 and DSD-5, respectively. The error percentage here is ob- and high levels (+). The factor ranges are selected to be broad and
FE − PTest
Pmax max FE Test practical based on previous experimental studies (e.g., [4,22]).
tained from Test × 100 , where Pmax and Pmax are the peak strength
Pmax
from the finite element analysis and the experimental test, respectively.
The accuracy of the finite element analysis in predicting the limit 3.3. Design of experiment
states of butterfly-shaped fuses is also verified against three specimen
models analyzed by Farzampour [28,32]. The detailed dimensions of The sensitivity analysis is carried out using the Design-Expert soft-
the models experiencing flexural yielding, shear yielding, and lateral ware [33]. A two-level fractional factorial design is considered for the
torsional buckling are listed in Table 1. These models include only one analysis. A full factorial design with seven factors and two levels for
link of the butterfly-shaped and slit fuses to reduce the computational each factor requires analyzing 27 (=128) models, which is computa-
cost. As the presented study is also considering one link of the fuse, the tionally expensive. Therefore, a one-fourth fractional factorial design
finite element model is validated with the reference numerical results. with a resolution of four has been employed in this study. By doing this,
803
A. Siar Mahmood Shah and S. Moradi Structures 23 (2020) 799–811
Fig. 8. Lateral load-displacement response of specimens DSD-5 and PSD-5: numerical results in comparison with experimental results [22].
(a) (b)
(c)
Fig. 9. Validation of the finite element modeling against a numerical study of three limit states [28,32]: (a) flexural yielding; (b) shear yielding; (c) lateral torsional
buckling.
804
A. Siar Mahmood Shah and S. Moradi Structures 23 (2020) 799–811
Table 3 Table 4
Factor combination for each model. Response characteristics of each model.
Model L : Length a: b: t : Thickness Fy : Steel E: c: Model Initial Yield Ultimate Effective Maximum Ductility
Mid- End- yield Modulus Band Stiffness Strength Stiffness Damping Strength
width Width strength of zone kN/mm kN kN/mm kN
elasticity length
mm mm mm mm MPa MPa mm 1 572.97 212.0 5.19 2600.0 565 107.03
2 959.38 307.0 2.04 2310.0 663 106.88
1 180 36 200 40 245 185,000 75 3 18.070 30.90 2.02 43,600 202 76.842
2 180 36 200 40 340 185,000 5.0 4 79.333 11.90 0.27 346.00 26.0 86.000
3 180 36 36.0 40 340 212,000 75 5 1.6887 8.680 0.40 10,400 50.7 19.260
4 180 15 200 5.0 245 185,000 5.0 6 0.1851 1.090 0.03 354.00 2.03 3.9728
5 500 15 36.0 40 245 212,000 5.0 7 54.545 24.00 0.37 14,300 131 200.45
6 500 15 36.0 5.0 245 185,000 5.0 8 2.6459 13.60 0.18 55,000 99.0 36.070
7 180 36 36.0 40 245 212,000 5.0 9 3.9444 4.260 0.09 344.00 6.97 11.944
8 500 36 36.0 40 340 185,000 5.0 10 1.8278 2.760 0.03 330.00 4.86 11.818
9 180 15 36.0 5.0 340 212,000 5.0 11 5.8000 3.190 0.11 646.00 6.86 32.727
10 180 15 36.0 5.0 245 212,000 75 12 510.71 143.0 5.19 1420.0 405 102.85
11 180 36 36.0 5.0 245 185,000 5.0 13 134.29 37.60 0.73 550.00 63.0 64.286
12 180 15 200 40 340 212,000 75 14 0.3730 1.750 0.02 962.00 3.80 7.2920
13 180 36 200 5.0 340 212,000 5.0 15 0.2412 1.240 0.05 78.300 1.89 2.5097
14 500 36 36.0 5.0 340 212,000 5.0 16 1.3398 12.30 0.06 11,300 59.5 11.372
15 500 36 36.0 5.0 245 212,000 75 17 55.312 17.70 1.25 34.200 25.2 10.469
16 500 15 36.0 40 340 212,000 75 18 15.000 18.30 0.82 173.00 28.8 10.574
17 180 15 200 5.0 340 185,000 75 19 11.590 27.70 0.64 111.00 38.5 5.3975
18 500 36 200 5.0 245 185,000 5.0 20 11.229 13.70 0.30 29.300 18.1 3.8442
19 500 36 200 5.0 340 185,000 75 21 12.748 21.80 0.67 4280.0 58.5 29.474
20 500 15 200 5.0 340 212,000 5.0 22 66.481 71.80 1.15 3020.0 198 41.667
21 180 15 36.0 40 245 185,000 75 23 706.67 106.0 2.04 1190.0 327 192.00
22 500 15 200 40 245 185,000 75 24 1.9037 4.550 0.08 642.00 6.97 9.7908
23 180 15 200 40 245 212,000 5.0 25 79.344 96.80 0.13 3260.0 283 72.295
24 180 36 36.0 5.0 340 185,000 75 26 136.06 166.0 2.59 13,100 372 63.442
25 500 15 200 40 340 185,000 5.0 27 1.6768 11.00 0.37 17,300 62.0 17.561
26 500 36 200 40 245 212,000 5.0 28 106.13 225.0 1.57 14,800 490 39.056
27 500 36 36.0 40 245 185,000 75 29 0.1488 1.690 0.06 139.00 2.68 2.0617
28 500 36 200 40 340 212,000 75 30 27.705 33.80 0.39 19,600 106 76.721
29 500 15 36.0 5.0 340 185,000 75 31 83.125 26.60 0.38 394.00 46.4 40.133
30 180 15 36.0 40 340 185,000 5.0 32 9.6226 10.20 1.18 4.5300 12.8 2.2547
31 180 36 200 5.0 245 212,000 75
32 500 15 200 5.0 245 212,000 75
design factors on the cyclic response characteristics of fuses.
Important factors and interactions that have significant effects on
negative peak displacements under the load-displacement plot, re- the cyclic response are identified by a half-normal probability plot. This
spectively. method is based on the theory that insignificant factors fall on a straight
( |P +| + |P −|) line and are normally distributed with a near-zero mean value [34].
keff = After generating the half-normal probability plot for each response
(|δ +| + | δ−|) (1)
variable, we obtain the effects of main factors and their interactions.
2 Eloop In order to check the significance of the statistical model for each
βeff = ⎛ ⎞ response variable, statistical analysis of variance (ANOVA) using the
π (k
⎝ ⎠ eff ) ∗ (|δ +| + | δ−|)2 (2)
Design-Expert software is performed [33]. The results from half-normal
Initial stiffness (K ) is calculated as the ratio of yield strength to yield probability plots are also confirmed. ANOVA is a method to statistically
displacement while ultimate stiffness (K ) is calculated from the slope of test the null hypothesis of no treatment effects [34]. A significance level
the last two steps of the response curve. Further, the ductility is ex- of 5% is chosen for this ANOVA study. Hence, the most significant
pressed as the ratio of maximum displacement to yield displacement. factors have p-values less than 0.05. For instance, considering the initial
Lastly, maximum load (P ) is considered as the maximum strength of the stiffness response characteristic, the significance level of the factors is
fuse, while the first occurrence of yielding is taken as the yield strength determined based on their p-values. The p-value for the fuse end-width
(Py ). (b ), thickness (t ), and length (L ) is less than 0.0001, which indicates the
significance of their influence on the initial stiffness response.
3.4. Results of the sensitivity analysis In order to ensure the high accuracy of the statistical model, R-
Squared, adjusted R-Squared, predicted R-Squared, and adequate pre-
The load-displacement cyclic response of each model is recorded. A cision are checked. R-Squared value is used to determine the percentage
wide range of cyclic response is observed. For example, the cyclic re- of the data that can be explained by the model. An R-Squared value of
sponse plots for three models 1, 7, and 12 are illustrated in Fig. 10. at least 0.80 is typically desired for a good model. The adjusted R-
Model 7 with the factor combinations shown in Table 3 has shown Squared and predicted R-Squared are used to define the relation of
greater ductility compared to models 1 and 12. The obtained results for experimental data to their predicted values of the final equation. The
the response characteristics of these three models show that the effec- difference between the adjusted R-Squared and the predicted R-Squared
tive damping in model 7 is respectively 5.50 times and 10 times higher should not be greater than 0.20. Adequate precision measures the signal
than those in models 1 and 12. The maximum strength of model 1 is to noise ratio and a value of greater than 4 is desirable [33]. For ex-
4.30 and 1.40 times greater than that in models 7 and 12, respectively. ample, considering the initial stiffness response, the value of R-Squared,
Lastly, the yield strength response is considerably higher in model 1 the difference of adjusted and predicted R-Squared, and the adequate
when compared to model 7 and 12. The sensitivity analysis results will precision are 0.93, 0.05, and 17.84, respectively.
be subsequently presented along with the discussions of the effect of the Percentage contributions of important factors and interactions for
805
A. Siar Mahmood Shah and S. Moradi Structures 23 (2020) 799–811
(a) (b)
(c)
Fig. 10. Hysteretic response plots for: (a) model 1; (b) model 7; (c) model 12.
each response characteristic are presented in Fig. 11. The positive sign stiffness are all present in the previously derived design formula (Eq.
indicates that the corresponding factor or interaction has a positive (3)). The equation further supports the positive effect of end-width (b )
effect. and thickness (t ) as well as the negative effect of length (L ), indicating
that the initial stiffness of the fuse is directly proportional to the end-
width and thickness and it is inversely proportional to the length.
3.5. Sensitivity of response characteristics
The 3D plot in Fig. 12 shows the effect of interaction (bt ) on the
initial stiffness. It is observed that the effect of the thickness (t ) on the
3.5.1. Initial stiffness
initial stiffness is significantly increased as the end-width (b ) is in-
In this section, the sensitivity of the initial stiffness response of steel
creased. As the thickness (t ) is increased (from t = 5 mm to 40 mm) at
plate fuses is discussed. As shown in Fig. 11(a), the most important
the lower level of the end-width (b = 36 mm), there is a slight change
factors considering the initial stiffness are the end-width (b ), thickness
in the initial stiffness of the plate (approximately from 0 to 10 kN/mm).
(t ), length (L ), as well as the interaction between end-width and
However, at the high level of the end-width (b = 200 mm), with the
thickness (bt ), length and end-width (Lb ), length and thickness (Lt ),
increase in the thickness (from t = 5 mm to t = 40 mm), the initial
and mid-width and modulus of elasticity (aE ) with the percentage
stiffness is significantly increased from 0 to approximately 400 kN/mm.
contributions of 22%, 15%, 15%, 13%, 12%, 9% and 7%, respectively.
The remaining factors shown in Fig. 11(a) contribute less to the initial
stiffness. 3.5.2. Yield strength
Among the significant factors, the end-width (b ), thickness (t ) and Another response value that is studied from the results of sensitivity
their interaction (bt ) show positive effects (i.e., the initial stiffness re- analysis is the yield strength of the steel plates. The yield strength is a
sponse is directly proportional to these factors). For example, by in- significant response characteristic in the steel plate fuses since the fuse
creasing the thickness from the low level (5 mm) to the high level should have enough yielding capacity to stay undamaged under service
(40 mm), the initial stiffness of the fuse is increased. On the other hand, loads [22]. The sensitivity analysis results show that the yield strength
the length (L ) and its interaction with the thickness (Lt ), its interaction of steel plate fuses is mostly affected by the end-width (b ), thickness (t ),
with the end-width (Lb ), and the interaction between the mid-width mid-width (a ), length (L ), as well as the interactions bt and ab having
and modulus of elasticity have negative effects on the initial stiffness percentage contributions of 29%, 29%, 5%, 2%, 18%, and 5%, re-
(aE ). This observation is also supported by the results of previous re- spectively as shown in Fig. 11(b). Among all the main factors, only the
search [4,35] reporting that the initial stiffness of the fuse is directly fuse length (L ) has a negative impact on the yield strength. In addition,
proportional to the end-width and thickness; and this response is in- it can be observed that over 75% of the variability in the fuse yield
versely proportional to the length. The formula for calculating the in- strength is controlled by the end-width (b ), thickness (t ) and their in-
itial stiffness is as follows: teraction (bt ). The analysis results are additionally confirmed by the
yield strength design formula (Eq. (4)) for slit fuses [4,31].
2Ebt 3
K=
3L3 (3) nb2tFy
Py =
3L (4)
It is observed that, the most significant factors influencing the initial
806
A. Siar Mahmood Shah and S. Moradi Structures 23 (2020) 799–811
(a) (b)
(c) (d)
(e) (f)
Fig. 11. Percentage contribution of important factors and interactions for each response characteristic.
It is perceived that the factors influencing the yield strength are all width (b ), thickness (t ), length (L ), and the interaction between length
present in the previously derived design formula (Eq. (4)). It should be and thickness (Lt). This observation is likewise the sensitivity results for
noted that Eq. (4) is derived only for straight (slit) fuses excluding the the maximum and yield strength, which were found sensitive to b , t ,
effect of mid-width (a ), which is included in the current sensitivity and L . The calculated percentage contributions for the main factors and
study. The equation further confirms the positive effect of end-width (b ) interactions indicate that these effects (i.e. the effect of b , t , L , c , Lt , bt ,
and thickness (t ) as well as the negative effect of length (L ) – showing Lc , and aE ) account for 24%, 20%, 7%, 5%, 8%, 6%, 4% and 3% of the
that the yield strength of the fuse is directly proportional to the end- total response variability, respectively. As illustrated in Fig. 11(c), end-
width and thickness and it is inversely proportional to the length. width (b ), thickness (t ) band zone (c ), and interaction (bt ) have positive
The 3D plot in Fig. 13 shows the interaction between the end-width effects while the remaining factors have negative effects.
(b ) and the thickness (t ) of the steel plate fuse. It can be observed that To determine the effect of interaction between length (L ) and
increasing both the end-width (b ) and the thickness (t ) from their low thickness (t ) on the variability of the ultimate stiffness response, the 3D
levels (t = 5 mm, b = 36 mm) to high levels (t = 40 mm, b = 200 mm) plot in Fig. 14 is illustrated. It can be observed that the ultimate stiff-
increases the yield strength from 0 to approximately 165 kN. ness response increases approximately from 0.40 to 2.25 kN as the
thickness (t ) increases from its low level (t = 5 mm) to high level
(t = 40 mm) and the length (L ) decreases from its high level
3.5.3. Ultimate stiffness (L = 500 mm) to its low level (L = 180 mm).
The ultimate stiffness of steel plate fuses is mostly affected by end-
807
A. Siar Mahmood Shah and S. Moradi Structures 23 (2020) 799–811
steel yield strength (bFy ). The remaining factors such as the length (L ),
modulus of elasticity (E ) and the band zone length (c ) have shown no
significant effect on the effective damping response characteristics.
From the significant factors, the thickness (t ), mid-width (a ), steel
yield strength (Fy ), the interactions between mid-width and thickness
(at ), thickness and steel yield strength (tFy ) have positive effects with
the percentage contributions of 29%, 8%, 5%, 7%, and 5%, respectively
as shown in Fig. 11(d). On the other hand, the end-width (b ), the in-
teractions between the end-width and thickness (bt ), and between the
end-width and modulus of elasticity (bE ) show negative effects on the
effective damping with percentage contributions of 12%, 11%, and 4%,
respectively.
These results indicate that the effective damping of steel plate fuses
is mainly controlled by the thickness, end-width, and their interaction
having a combined contribution of more than 50%. This trend can be
attributed to the fact that the energy dissipating capacity of steel plate
Fig. 12. The effect of interaction between the thickness and the end-width on fuse is increased significantly with an increase of the thickness because
initial stiffness. of the delay in the buckling of the fuses. This is further supported by
Farzampour and Eatherton [26], suggesting that the thickness of the
plate is observed to control the buckling limit state of the fuse, and thus
affecting the energy dissipation capability. Further, the negative effect
of the end-width on the effective damping is expected because in-
creasing the end-width over the length causes the fuse to act more like a
solid plate thus leading to less energy dissipation. These observations
are also consistent with the research by Farzampour and Eatherton
[26]. They reported that increasing the end width over the length of the
butterfly-shaped fuses would make the shear mechanism governing
over the flexure, and thus resulting in a reduction in the energy dis-
sipation [26].
To examine the effect of the interaction between the main factors
end-width (b ) and thickness (t ) on the variability of the effective
damping of steel plate fuses, the 3D plot in Fig. 15 is presented. It can be
observed that at the low level of end-width (b = 36 mm), increasing the
thickness (t ) significantly increases the effective damping response (1 to
215,000). In contrast, at the high level of end-width (b = 200), the
effective damping is increased at a lesser rate (from 1 to 5000) due to
Fig. 13. The effect of interaction between the end-width and the thickness on the increase in the thickness (t ). This trend is because of the active
yield strength.
interaction between b and t . In other words, since the interaction is
involved, the amount of change in the response because of varying
thickness (t ) is different at different levels of the end-width (b ).
Fig. 14. The effect of interaction between the length and the thickness on ul-
timate stiffness.
808
A. Siar Mahmood Shah and S. Moradi Structures 23 (2020) 799–811
• The fuse link end-width and thickness significantly influence all the
response characteristics. The fuse length and mid-width are influ-
Fig. 16. The effect of interaction between the end-width and thickness on the ential on some of the response variables.
maximum strength. • By varying modulus of elasticity (from 185,000 MPa to
809
A. Siar Mahmood Shah and S. Moradi Structures 23 (2020) 799–811
212,000 MPa), steel yield strength (from 245 MPa to 340 MPa), and Acknowledgment
the band zone length (from 5 mm to 75 mm), it is observed that the
cyclic performance of steel plate fuses is not significantly affected by The financial support from the Natural Sciences and Engineering
the variation in these factors. The sensitivity analysis shows that the Research Council of Canada (NSERC) through Discovery Grant and the
variability of modulus of elasticity, steel yield strength, band zone in Ryerson University Faculty of Engineering and Architectural Science is
the response quantities is generally below 1% except for the effec- gratefully acknowledged.
tive damping and ductility having percentage contributions close to
or less than 5%. References
• The initial stiffness of a steel plate fuse is mostly influenced by the
end-width, thickness, length and their interactions. These para- [1] Enrique Martı́nez-Rueda Juan. On the evolution of energy dissipation devices for
meters account for approximately 92% of the total variability in the seismic design. Earthq Spectra 2002;18:309–46.
[2] Vargas R, Bruneau M. Experimental response of buildings designed with metallic
initial stiffness response. Further, the length and its interactions structural fuses. II. J Struct Eng 2009. https://doi.org/10.1061/(ASCE)0733-
with the thickness and end-width have a negative effect on the in- 9445(2009) 135:4(394).
itial stiffness of steel plate fuses. [3] Deierlein G, Krawinkler H, Ma X, Eatherton M, Hajjar J, Takeuchi T, et al.
• The factors that mostly influence the load capacity of steel plate
Struct Des Tall Build 1998:21–32. https://doi.org/10.1002/(SICI)1099-
1794(199803)7:1<21::AID-TAL105>3.0.CO;2-8.
fuses are the thickness, end-width, mid-width, and length. These [13] Lynch JP. Active Structural Control Research at Kajima Corporation 1998:1–53.
factors with their interactions account for nearly 94% of the total [14] Onishi K, Fukuzawa E, Fukumoto T. Structural design and construction work of
Kamogawa grand tower using steel damper. Concr J 1993;31:28–38. https://doi.
variability in the response. Among these, over 80% of the load ca-
org/10.3151/coj1975.31.10_28.
pacity variability is controlled by the thickness, end-width and their [15] Kareem A, Kijewski T, Tamura Y. Mitigation of motions of tall buildings with
interactions. Further, the length and its interaction with thickness specific examples of recent applications. Wind Struct An Int J 1999.
shows a negative effect on the load capacity response while all the [16] Luth GP, Sargunuraj S, Krawinkler H, McDonald B. USC school of cinema an ex-
ample of reparable performance based design, Menlo Park, California: SEAOC 2008
other factors have positive effects. In addition, with increasing the CONVENTION PROCEEDINGS; 2008, p. 1–20.
end-width for thick fuses, the load capacity is increased sig- [17] Li HN, Li G. Earthquake-resistant design of rc frame with “dual functions” metallic
nificantly. dampers. Am Soc Mech Eng Press Vessel Pip Div PVP 2008. https://doi.org/10.
•
1115/PVP2007-26450.
The most important factors and interactions influencing the ductility [18] Tagami J, et al. Low cycle fatigue test of steel plate dampers with honeycomb
of steel plate fuses are the thickness, length, end-width, the band shaped openings. Chugoku, Japan: Summ. Tech. Pap. Annu. Meet. AIJ; 1999. p.
zone length, as well as some interactions that have less than 5% 791–6.
[19] El-Bahey S, Bruneau M. Structural fuse concept for bridges. 9th US Natl. 10th Can.
contributions. Overall, these factors and interactions account for Conf. Earthq. Eng. 2010, Incl. Pap. from 4th Int. Tsunami Symp., 2010.
about 85% of the total variability in the response. Among these [20] El-Bahey S, Bruneau M. Bridge piers with structural fuses and bi-steel columns. II:
important factors, the length and thickness are found to control over analytical investigation. J Bridg Eng 2012;17:36–46. https://doi.org/10.1061/
(ASCE)BE.1943-5592.0000236.
50% of the ductility response variability. Further, it was observed [21] El-Bahey S, Bruneau M. Bridge piers with structural fuses and bi-steel columns. I:
that shorter and thicker fuses possess greater ductility. experimental testing. J Bridg Eng 2012;17:25–36. https://doi.org/10.1061/(ASCE)
BE.1943-5592.0000234.
[22] Lee CH, Ju YK, Min JK, Lho SH, Kim SD. Non-uniform steel strip dampers subjected
This study and further complementary research contribute to the
to cyclic loadings. Eng Struct 2015;99:192–204. https://doi.org/10.1016/j.
efficient, robust, and reliable design of energy dissipating steel plate engstruct.2015.04.052.
fuses, which are key elements in high-performance earthquake resistant [23] Lee CH, Lho SH, Kim DH, Oh J, Ju YK. Hourglass-shaped strip damper subjected to
structures. The results from this study can be used to develop predictive monotonic and cyclic loadings. Eng Struct 2016;119:122–34. https://doi.org/10.
1016/j.engstruct.2016.04.019.
equations for the response of steel plate fuses. In addition, with the [24] Farzampour A, Eatherton M. Parametric study on butterfly- shaped shear links with
identified significant factors, optimization studies can be performed to various geometries. 11th US Natl Conf Earthq Eng 2018;24060.
determine the optimal design of energy dissipating steel plate fuses. [25] Koppal M, Eatherton MR. Perforated steel plate shear walls for tunable seismic
resistance. Struct. Congr. 2013 Bridg. Your Passion with Your Prof. - Proc. 2013
Struct. Congr., 2013.
[26] Farzampour A, Eatherton MR. Yielding and lateral torsional buckling limit states for
Declaration of Competing Interest butterfly-shaped shear links. Eng Struct 2019;180:442–51. https://doi.org/10.
1016/j.engstruct.2018.10.040.
[27] Teruna DR, Majid TA, Budiono B. Experimental study of hysteretic steel damper for
The authors declare that they have no known competing financial energy dissipation capacity. Adv Civ Eng 2015:1–12. https://doi.org/10.1155/
interests or personal relationships that could have appeared to influ- 2015/631726.
[28] Farzampour A. Evaluating shear links for use in seismic structural fuses. Virginia
ence the work reported in this paper.
810
A. Siar Mahmood Shah and S. Moradi Structures 23 (2020) 799–811
811